AI training Switzerland

Start building your personal staff.

Everyone deserves a small team that does the boring work. By the end of one workshop, you have one — and one less thing on your list.

We pick something you wish you didn’t have to do, and we build the AI staff member who does it for you. No jargon. No sandbox theatre. Live, in your tools, with your data.

For people, not engineers
One workflow per workshop
Live build, not slides
You leave with it running

Your personal staff

Six small hires you wouldn’t miss.

Each one is one task off your week. Pick one to start. Add the rest as you go.

Inbox triage

Sorts your morning email by urgency and drafts the boring replies.

Meeting clerk

Summarises the call, lists the actions, drafts the follow-up.

Document drafter

Quotes, briefs, contracts, reports — first draft from your data.

Spreadsheet cleaner

Tidies messy exports, cross-checks numbers, surfaces anomalies.

Research assistant

Reads ten tabs so you don't have to. Comparisons, summaries, sources.

Personal admin

Schedules appointments, plans groceries, sorts receipts, books travel.

Mix personal and work freely. The same workshop can build the spreadsheet cleaner for Monday and the grocery planner for Sunday.

Plain answers

Some of the common questions we answer.

What is an agent, exactly?

A prompt running in a loop. The model takes text in (your goal, plus a list of tools it's allowed to call), writes text out (a request to call a tool, or a final answer), we run the tool, paste the result back in, and ask the model again. It loops until the model stops asking for tools. The intelligence is just next-token prediction. The loop is what turns prediction into work.

Is it safe?

As safe as the tools you wire to it. The model itself is text-in, text-out — it can only do what your tool list lets it do. Risk lives in three places: read vs write permissions, sensitive vs public data, and what gets auto-approved versus human-reviewed. We start read-only, drafts only (no sends), one workflow at a time. Every model call is logged so you can see what it asked, what it got back, and what it did.

Does it take hours to set up?

The agent itself is minutes. Wiring it to your inbox, calendar, CRM and getting auth through IT is an hour or two. Writing the prompt that produces good output reliably is anywhere from an afternoon to a couple of weeks — that's where 80% of the work sits. The first one is slow. By the fourth, you're cloning patterns and most of the time goes into data plumbing, not the agent.

Can it do my grocery shopping?

The software part — yes: read what's in the fridge, plan a week of meals, generate the list, call your supermarket's online ordering API. The physical part — no: it can't pick up bags. The line is wherever digital ends and atoms begin. Until you wire it to a robot with hands, it stops at 'order placed.'

Do I need to be technical?

To use one, no. To configure one well, yes — but 'technical' here means clear thinking, not code. The hard parts are deciding what good output looks like, picking which steps a human reviews versus auto-runs, and naming the edge cases. We do that in workshops with the person who has the boring task, not their developer. The tooling does the typing.

What if it gets it wrong?

It will, regularly. Models hallucinate — they state things confidently that aren't true. Tools return errors the model misreads. Prompts hit cases nobody anticipated. The defence isn't 'make it perfect' — it's review gates and small blast radius. Every action is drafted (you check before send), reversible (you can undo it), or audited (you can trace it later). Set those right and wrong becomes caught, not damage done.

How the workshop works

Two hours. One concrete task. One new staff member.

We don’t teach AI in the abstract. We sit down with the task you brought, set up the tools, and build the agent in front of you. You leave with it running.

1
Tell us the task

Five-minute intake: what you do today, what you wish you didn’t.

2
We pick the right setup

Claude, ChatGPT, Codex, Zapier, a small Salesforce tweak — the lightest tool that gets the job done safely.

3
We build it together

Live, in your tools, with your real data. You watch every decision so you can change it later.

4
You take it home

Your new staff member runs from your account. Plus a 1-page playbook so you can hire the next one.

The honest version

You can’t outsource judgement. You can outsource boredom.

We don’t pretend AI replaces the parts of your job that need taste or decisions. We make sure the parts that don’t — the typing, the sorting, the copy-pasting, the chasing — stop eating your week.

What would you stop doing tomorrow if you had a junior?

Bring us that task. Two hours from now you have a junior who does it. Personal, work, or both — we don’t mind which.